Explore the scholarly record.

Abstract

Autonomous vehicles (AVs) are increasingly interacting with human-driven vehicles (HVs), creating safety problems in which private information, human adaptation and technological penetration interact. Existing transportation studies have examined Signaling games, evolutionary AV–HV interaction, vehicle-to-vehicle (V2V) information and bounded rationality largely as separate mechanisms. This study develops a coupled Signaling–evolutionary safety model in which an AV privately observes its safety type, chooses whether to transmit an authenticated safety signal, and interacts with human drivers whose responses depend on the observed signal and evolve over repeated encounters. The AV penetration rate is modelled separately from the prior probability of a high-safety AV type and enters the behavioral and Signaling incentives through a penetration-dependent collision-risk structure. The static analysis derives necessary and sufficient incentive conditions for a strict separating Perfect Bayesian Equilibrium. Informative separation requires a type-specific signal-cost window, LAδLp<c<LAδHp , together with receiver conditions determining whether a signal can induce defensive driving. These conditions yield a closed-form lower penetration boundary. The static model is then embedded in a logit-regularised evolutionary system describing signal-conditioned human responses and type-specific Signaling adaptation. Local stability is evaluated from the spectral abscissa of the resulting four-dimensional Jacobian. Under an illustrative parameterisation, the receiver threshold is pR=0.6944 , and strict separation at full penetration is feasible for 0.20<c<0.72 . At p=0.90 and c=0.40 , the static separating conditions hold. The associated regularised dynamics exhibit both a stable near-separating state and an unstable stationary state, demonstrating that incentive compatibility does not by itself guarantee dynamic robustness. The results provide a tractable framework for evaluating AV communication design jointly with deployment conditions and human behavioral adaptation.